Traffic Light Silhouettes Recognition using Fourier Descriptors

P. Foucher, D. Moreno Eddowes, and J. Lopez Krahe (France)


Image processing, Fourier descriptors, visual handicapped.


This paper presents a study of a new tool to recognize the pedestrian traffic lights. The use of mobile, usual and autonomous system is the basis of our approach. The traffic lights interpretation is based on the analysis of the images of silhouette placed beside the user (on the same side of the street). After a brief description of the objects extraction methods, this paper introduces the shape recognition tool based on the Fourier descriptors. The Fourier descriptors make possible to encode the boundary of the shape. This encoding can be compared with a model and a diagnosis concerning the traffic light identification which can be emitted. The Fourier descriptors have been tested and analysed on 86 images (44 green silhouettes and 42 red silhouettes). The Fourier descriptors give us very encouraging results and this approach could be used in the medium-term in a known environment.

Important Links:

Go Back